News Summary

• The u.s. government has banned access to Anthropic's Fable and Mythos AI models.
• The company says the move is in response to a security breach.
• Anthropic says the breach involved a method of bypassing guardrails.
• The jailbreak appears to have been reported by Amazon, which is both an investor in Anthropic and a major provider of inference to the company.
• A world where models are interchangeable is one where their differentiation is fleeting, and costs never cover their costs.